Merge pull request #104 from lpiwowar/lpiwowar/rag

Configure OGX with RAG

Author: openshift-merge-bot[bot]

Makefile

@@ -274,6 +274,8 @@ $(KUTTL): $(LOCALBIN)
 
 .PHONY: kuttl-test
 kuttl-test: kuttl ## Run kuttl tests
+	@command -v diff >/dev/null 2>&1 || { echo "ERROR: 'diff' command is required for KUTTL tests but not found in PATH" >&2; exit 1; }
+	@command -v oc >/dev/null 2>&1 || { echo "ERROR: 'oc' command is required for KUTTL tests but not found in PATH" >&2; exit 1; }
 	$(LOCALBIN)/kubectl-kuttl test --config kuttl-test.yaml test/kuttl/tests $(KUTTL_ARGS)
 
 .PHONY: kuttl-test-run

api/v1beta1/openstacklightspeed_types.go

@@ -23,8 +23,11 @@ import (
 )
 
 const (
+	// TODO(lpiwowar): Replace this with a stable (non-alpha) image version once
+	// the automated pipeline for building OGX-compatible vector database images
+	// is ready.
 	// OpenStackLightspeedContainerImage is the fall-back container image for OpenStackLightspeed
-	OpenStackLightspeedContainerImage = "quay.io/openstack-lightspeed/rag-content:os-docs-2025.2"
+	OpenStackLightspeedContainerImage = "quay.io/openstack-lightspeed/rag-content:alpha-ogx-os-docs-2025.2"
 
 	// LCoreContainerImage is the fall-back container image for LCore
 	LCoreContainerImage = "quay.io/lightspeed-core/lightspeed-stack:latest"

bundle/manifests/openstack-lightspeed-operator.clusterserviceversion.yaml

@@ -25,7 +25,7 @@ metadata:
       ]
     capabilities: Basic Install
     categories: AI/Machine Learning
-    createdAt: "2026-04-09T11:08:52Z"
+    createdAt: "2026-05-17T15:20:36Z"
     description: AI-powered virtual assistant for Red Hat OpenStack Services on OpenShift
     features.operators.openshift.io/cnf: "false"
     features.operators.openshift.io/cni: "false"
@@ -255,7 +255,7 @@ spec:
                     fieldRef:
                       fieldPath: metadata.annotations['olm.targetNamespaces']
                 - name: RELATED_IMAGE_OPENSTACK_LIGHTSPEED_IMAGE_URL_DEFAULT
-                  value: quay.io/openstack-lightspeed/rag-content:os-docs-2025.2
+                  value: quay.io/openstack-lightspeed/rag-content:alpha-ogx-os-docs-2025.2
                 - name: RELATED_IMAGE_LCORE_IMAGE_URL_DEFAULT
                   value: quay.io/lightspeed-core/lightspeed-stack:latest
                 - name: RELATED_IMAGE_EXPORTER_IMAGE_URL_DEFAULT
@@ -433,7 +433,7 @@ spec:
     name: Red Hat
     url: https://github.com/openstack-lightspeed/operator
   relatedImages:
-  - image: quay.io/openstack-lightspeed/rag-content:os-docs-2025.2
+  - image: quay.io/openstack-lightspeed/rag-content:alpha-ogx-os-docs-2025.2
     name: openstack-lightspeed-image-url-default
   - image: quay.io/lightspeed-core/lightspeed-stack:latest
     name: lcore-image-url-default

config/manager/manager.yaml

@@ -70,8 +70,11 @@ spec:
             valueFrom:
               fieldRef:
                 fieldPath: metadata.namespace
+            # TODO(lpiwowar): Replace this with a stable (non-alpha) image version once
+            # the automated pipeline for building OGX-compatible vector database images
+            # is ready.
           - name: RELATED_IMAGE_OPENSTACK_LIGHTSPEED_IMAGE_URL_DEFAULT
-            value: quay.io/openstack-lightspeed/rag-content:os-docs-2025.2
+            value: quay.io/openstack-lightspeed/rag-content:alpha-ogx-os-docs-2025.2
           - name: RELATED_IMAGE_LCORE_IMAGE_URL_DEFAULT
             value: quay.io/lightspeed-core/lightspeed-stack:latest
           - name: RELATED_IMAGE_EXPORTER_IMAGE_URL_DEFAULT

hack/env.sh

@@ -2,5 +2,8 @@
 export RELATED_IMAGE_LCORE_IMAGE_URL_DEFAULT="quay.io/lightspeed-core/lightspeed-stack:latest"
 export RELATED_IMAGE_EXPORTER_IMAGE_URL_DEFAULT="quay.io/lightspeed-core/lightspeed-to-dataverse-exporter:latest"
 export RELATED_IMAGE_POSTGRES_IMAGE_URL_DEFAULT="registry.redhat.io/rhel9/postgresql-16:latest"
-export RELATED_IMAGE_OPENSTACK_LIGHTSPEED_IMAGE_URL_DEFAULT=quay.io/openstack-lightspeed/rag-content:os-docs-2025.2
+# TODO(lpiwowar): Replace this with a stable (non-alpha) image version once
+# the automated pipeline for building OGX-compatible vector database images
+# is ready.
+export RELATED_IMAGE_OPENSTACK_LIGHTSPEED_IMAGE_URL_DEFAULT="quay.io/openstack-lightspeed/rag-content:alpha-ogx-os-docs-2025.2"
 export WATCH_NAMESPACE="openstack-lightspeed"

internal/controller/assets/vector_database_build.py

@@ -0,0 +1,291 @@
+#!/usr/bin/env python3
+
+#
+# Copyright 2026.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Vector database configuration builder for the OpenStack Lightspeed operator.
+
+Runs as the second init container (`vector-database-config-build`), after
+`vector_database_collect.sh`. It loads operator-provided base configs, walks every
+vector DB directory left by the collect step, and writes merged configs back to
+the shared volume.
+
+Input layout (under --vector-db-path, produced by vector_database_collect.sh):
+  {vector-db-path}/
+  └── <image_uuid_dir>/             (random directory name from collect script)
+      ├── vector_db/
+      │   ├── <vector-db-name>/
+      │   │   ├── llama-stack.yaml
+      │   │   └── faiss_store.db
+      │   └── ocp_X.YZ/             (optional, when OCP RAG is enabled)
+      │       ├── llama-stack.yaml
+      │       └── faiss_store.db
+      └── embeddings_model/
+
+Output (written to --vector-db-path, same basenames as the base configs):
+  {vector-db-path}/
+  ├── ogx_config.yaml
+  ├── lightspeed-stack.yaml
+  └── <collect-dir>/                (collected data preserved)
+
+Processing:
+  1. For each subdirectory of */vector_db/, read its llama-stack.yaml.
+  2. For each detected llama-stack.yaml file, extract its data and inject
+     the relevant entries into the output ogx_config.yaml and lightspeed-stack.yaml
+     files.
+  3. Write the merged YAML next to the collected data.
+
+Warning: This script only injects values into existing config structures.
+Base configs MUST contain otherwise this script will fail:
+- OGX config: registered_resources.{models,vector_stores}, storage.backends,
+              providers.{inference,vector_io}
+- Lightspeed Stack config: byok_rag, rag.inline
+
+Arguments:
+  --vector-db-path            Shared volume path (input collected data, output configs)
+  --ogx-config-path           Path to the base OGX configuration file
+  --lightspeed-stack-path     Path to the base Lightspeed Stack configuration file
+"""
+
+import argparse
+from pathlib import Path
+from typing import Any, Iterable, Optional, Callable
+import logging
+import sys
+
+import yaml
+
+# Template for the directory path where data for a single vector database
+# instance resides. In the configuration, VECTOR_DB_DATA_PATH will typically
+# be substituted by the operator as an environment variable.
+VECTOR_DB_DIR_TEMPLATE = (
+    "${{env.VECTOR_DB_DATA_PATH}}/{uuid}/vector_db/{vector_db_name}"
+)
+
+# Template for the directory path where data for an embedding model resides. In
+# the configuration, VECTOR_DB_DATA_PATH will be substituted by OGX using
+# an environment variable.
+EMBEDDING_MODEL_DIR_TEMPLATE = "${{env.VECTOR_DB_DATA_PATH}}/{uuid}/embeddings_model"
+
+# Template for a file path where vector db data are stored.
+VECTOR_DB_DATA_PATH_TEMPLATE = f"{VECTOR_DB_DIR_TEMPLATE}/faiss_store.db"
+
+# The original configuration file name for OGX in the mounted vector database data.
+# Update later: The file is still named 'llama-stack.yaml' for backward compatibility,
+# since the Llama Stack project was renamed to OGX recently.
+OGX_CONFIG_SOURCE_FILE_NAME = "llama-stack.yaml"
+
+
+# -- Shared functions --------------------------------------------------------
+def load_yaml_file(yaml_file_path: Path) -> dict[str, Any]:
+    """Load YAML file"""
+    try:
+        with open(yaml_file_path, "r", encoding="utf-8") as f:
+            return yaml.safe_load(f) or {}
+    except FileNotFoundError:
+        logging.error("YAML file not found: %s", yaml_file_path)
+        sys.exit(1)
+
+
+def add_unique(lst: list, item: Any, key: Optional[str] = None) -> None:
+    """Add item to list if not already present.
+
+    :param lst:  List to modify in-place
+    :param item: Item to add
+    :param key:  If provided, check uniqueness by comparing item[key] values.
+                 If None, check direct item equality.
+    """
+    if key:
+        if any(existing.get(key) == item.get(key) for existing in lst):
+            return
+    elif item in lst:
+        return
+    lst.append(item)
+
+
+def write_yaml_file(yaml_data: dict[str, Any], dest_path: Path) -> None:
+    """Write YAML data to the specified file path."""
+    try:
+        dest_path.parent.mkdir(parents=True, exist_ok=True)
+        with open(dest_path, "w", encoding="utf-8") as f:
+            yaml.dump(yaml_data, f, default_flow_style=False, sort_keys=False)
+    except (OSError, yaml.YAMLError) as e:
+        logging.error("Failed to write YAML to %s: %s", dest_path, e)
+        sys.exit(1)
+
+
+def iterate_vector_db_data_dir(vector_db_data_dir_path: Path) -> Iterable[Path]:
+    """Return all folders inside any vector_db/ subfolder, one per yield."""
+    for image_uuid_dir in vector_db_data_dir_path.iterdir():
+        vector_db_path = image_uuid_dir.joinpath("vector_db")
+
+        if not vector_db_path.is_dir():
+            continue
+
+        for folder in vector_db_path.iterdir():
+            if folder.is_dir():
+                yield folder
+
+
+def config_build(
+    vector_db_parent_dir: Path,
+    config_target_path: Path,
+    config_populate_fn: Callable[[Path, dict[str, Any]], dict[str, Any]],
+) -> None:
+    config_target = load_yaml_file(config_target_path)
+    for vector_db_dir in iterate_vector_db_data_dir(vector_db_parent_dir):
+        ogx_config_source_path = vector_db_dir.joinpath(OGX_CONFIG_SOURCE_FILE_NAME)
+        try:
+            config_target = config_populate_fn(ogx_config_source_path, config_target)
+        except (KeyError, IndexError) as e:
+            logging.error(
+                "Error processing config: missing required section in source "
+                "or target file (%s)",
+                e,
+            )
+            sys.exit(1)
+
+    config_product_path = vector_db_parent_dir.joinpath(config_target_path.name)
+    write_yaml_file(config_target, config_product_path)
+
+
+# ----------------------------------------------------------------------------
+
+
+# -- OGX functions -----------------------------------------------------------
+def ogx_process(ogx_config_source_path: Path, ogx_config_target: dict[str, Any]):
+    """Populate the target OGX config with vector DB data from source OGX config"""
+    ogx_config_source = load_yaml_file(ogx_config_source_path)
+
+    # E.g.: /data/<uuid>/vector_db/os_product_docs/llama-stack.yaml -> <uuid>
+    image_uuid = ogx_config_source_path.parts[-4]
+
+    # E.g.: /data/<uuid>/vector_db/os_product_docs/llama-stack.yaml -> os_product_docs
+    vector_db_name = ogx_config_source_path.parts[-2]
+
+    vector_db_file = VECTOR_DB_DATA_PATH_TEMPLATE.format(
+        uuid=image_uuid, vector_db_name=vector_db_name
+    )
+    embedding_model_dir = EMBEDDING_MODEL_DIR_TEMPLATE.format(uuid=image_uuid)
+
+    # Populate registered_resources.models
+    src_model = ogx_config_source["registered_resources"]["models"][0].copy()
+    src_model["provider_model_id"] = embedding_model_dir
+    tgt_models = ogx_config_target["registered_resources"]["models"]
+    add_unique(tgt_models, src_model, "model_id")
+
+    # Populate registered_resources.vector_stores
+    embedding_model = f"{src_model['provider_id']}/{embedding_model_dir}"
+    src_vstore = ogx_config_source["registered_resources"]["vector_stores"][0].copy()
+    src_vstore["embedding_model"] = embedding_model
+    tgt_vstores = ogx_config_target["registered_resources"]["vector_stores"]
+    add_unique(tgt_vstores, src_vstore)
+
+    # Populate storage.backends
+    storage_backend_key = f"kv_rag_{image_uuid}_{vector_db_name}"
+    storage_backend = ogx_config_source["storage"]["backends"]["kv_rag"].copy()
+    storage_backend["db_path"] = vector_db_file
+    ogx_config_target["storage"]["backends"][storage_backend_key] = storage_backend
+
+    # Populate providers.inference
+    src_inference = ogx_config_source["providers"]["inference"][0]
+    tgt_inferences = ogx_config_target["providers"]["inference"]
+    add_unique(tgt_inferences, src_inference)
+
+    # Populate providers.vector_io
+    src_vector_io = ogx_config_source["providers"]["vector_io"][0].copy()
+    src_vector_io["config"]["persistence"]["backend"] = storage_backend_key
+    tgt_vector_ios = ogx_config_target["providers"]["vector_io"]
+    add_unique(tgt_vector_ios, src_vector_io)
+
+    return ogx_config_target
+
+
+# ----------------------------------------------------------------------------
+
+
+# -- Lightspeed Stack functions ----------------------------------------------
+def lstack_process(
+    ogx_config_source_path: Path, lstack_config_target: dict[str, Any]
+) -> dict[str, Any]:
+    """Update Lightspeed stack config with RAG entries from OGX config source."""
+    ogx_config_source = load_yaml_file(ogx_config_source_path)
+
+    src_vstores = ogx_config_source["registered_resources"]["vector_stores"]
+    vector_store_id = src_vstores[0]["vector_store_id"]
+
+    add_unique(
+        lstack_config_target["byok_rag"],
+        {
+            "rag_id": vector_store_id,
+            "vector_db_id": vector_store_id,
+            # The score multiplier is set to 1.0 so all BYOK sources have
+            # equal weighting.
+            "score_multiplier": 1.0,
+            # The Lightspeed Stack currently requires a "db_path" value even
+            # when OGX operates in server mode. This placeholder value ("NONE")
+            # is provided solely to satisfy this requirement and should be
+            # removed once the Lightspeed Stack no longer mandates it for
+            # server mode.
+            "db_path": "NONE",
+        },
+    )
+
+    add_unique(lstack_config_target["rag"]["inline"], vector_store_id)
+    return lstack_config_target
+
+
+# ----------------------------------------------------------------------------
+
+
+def parse_arguments() -> argparse.Namespace:
+    """Parse command-line arguments and return parsed namespace."""
+    parser = argparse.ArgumentParser(
+        description=(
+            "Build vector database configuration files by merging collected "
+            "vector DB data with base configs"
+        )
+    )
+    parser.add_argument(
+        "--vector-db-path",
+        type=Path,
+        required=True,
+        help="Path (as pathlib.Path) to the mounted vector DB data volume and output destination",
+    )
+    parser.add_argument(
+        "--ogx-config-path",
+        type=Path,
+        required=True,
+        help="Path (as pathlib.Path) to the base OGX configuration file",
+    )
+    parser.add_argument(
+        "--lightspeed-stack-path",
+        type=Path,
+        required=True,
+        help="Path (as pathlib.Path) to the base Lightspeed Stack configuration file",
+    )
+
+    return parser.parse_args()
+
+
+def main() -> None:
+    """main"""
+    args = parse_arguments()
+    config_build(args.vector_db_path, args.ogx_config_path, ogx_process)
+    config_build(args.vector_db_path, args.lightspeed_stack_path, lstack_process)
+
+
+if __name__ == "__main__":
+    main()